Bandwidth is the outgoing traffic from your Opensolr index to your site or app. You do not notice it until you run out. Three habits keep it low without making search slower.
Only the answer travels back to you, and only that is billed. Everything below makes it smaller or rarer.
01 · Cache results on your own side
Cache search results locally, with Memcached, Redis, or whatever your framework offers, so a page reload, an autocomplete keystroke or a back-button click does not hit Solr again. Fewer requests, faster pages, lower bandwidth.
02 · Return only what you need
Tune your /select requests with rows and fl so you fetch the records and the fields you actually display, and nothing else.
/select?q=mysearch&rows=10&fl=id,title
Audit your queries
Returning 100,000 records to display 10 is the single most common cause of a bandwidth bill. Look at the largest responses first.
Compress the responses
If your client supports it, enable gzip on the Solr responses. The same results, a fraction of the bytes.
Watch where it goes
Your control panel shows consumption per index. See how Solr traffic bandwidth is measured for what counts and what does not.
03 · Bandwidth is counted per index
Because the quota applies to each index, replication gives you more of it. Create index A, replicate it to index B, and have your application query both in round robin. Two indexes, two pools. See Solr replication for how the mechanism works.
Not sure which queries are costing you?
Send us the index name. We can look at the request log and tell you which query shapes are producing the traffic.